Your prompt… plus everything you've ever (publicly) written
TL;DR: Increasingly when you interact with frontier models, your prompt doesn't only contain the words on the page, but also a lot of what you've ever publicly written online.This is not intended to be a scary post.
The article discusses how modern AI models are increasingly able to recall specific details from users' online histories, which can influence their interactions with users.
What happened
The article explains that AI models are becoming better at recalling specific details from users' online activities, which can affect their interactions. It highlights examples where models recognized users based on their past writings and how this can influence the models' responses.
Why it matters
Understanding how AI models use users' online histories can help developers and users manage interactions more effectively, especially in professional or safety-critical contexts.
- AI models are increasingly capable of recalling specific details from users' online histories.
- Users should be aware that their past writings can influence AI interactions.
- Proactively addressing online history with AI models can improve co-working experiences.
The bigger picture
This awareness can lead to more transparent and effective interactions between users and AI models, particularly in scenarios where past user behavior might influence current outputs.
The article highlights the growing ability of AI models to recall user-specific details from online histories, emphasizing the importance of awareness and proactive management of these interactions.
Technical details
- Latent space fidelity: The ability of models to retain and recall specific details from training data.
- Attention mechanisms: Models may focus more on certain topics, like AI-related content, during training.
- Prompt influence: Users' past writings can shape the context and responses of AI models during interactions.
FAQ
How do AI models recall specific details from users' online histories?
AI models use their latent space, which is a compressed representation of training data, to recall specific details. Improved latent space fidelity allows models to remember nuanced information from users' past writings.
Why is it important for users to be aware of their online history influencing AI interactions?
Being aware helps users manage how AI models interact with them, especially in professional or safety-critical contexts where past user behavior might affect current outputs.
What can users do to address their online history with AI models?
Users should proactively address their online history with AI models by introducing themselves and clarifying their past writings, which can improve the effectiveness of interactions.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by LessWrong on Oct 14, 2026.
Comments
Sign in to join the discussion.